[EpiData-list] Design decision and information (EpIData Analysis)

epidata-list at lists.umanitoba.ca epidata-list at lists.umanitoba.ca
Sat Oct 7 09:12:38 CDT 2006


In the design of v2.0 of EpiData Analysis I wish to simplify and clarify
the use of options and "set" definitions to a unified approach. And this 
principle is more important
than backwards continuation of an unclear principle. The reason is among 
other
that v2.0 will include more statistics, e.g. gamma coefficient for 
ordinal data and
exact statistics for tables in general.

Another intention is to have more options and fewer commands, but to assist
users by more intuitive dialogs. E.g. as now where graph commands  can be
specified with a structured dialog where most options are available. The
intention is that :
v2.0: will include more functionality at the command level, but the 
dialogs helping
 beginners will come later.

The problem with the current way of implementation is that there is a 
non-unified
approach to what is defined at command level and what at "set" level. 
Mostly
due to the way it was done in Epi6. And this confuses users.

When the first release of v2.0 is ready all users will be able to 
comment on the
actual implementation of the principles. The release of v2.0 for test 
will be made
when we have reached a certain level of clarification. We are not quite 
there yet.
But good progress is being made.

I would like users to comment on these decisions:
.................................................................................................................
option:
A specification for a given command. These are a combination of / and 
letters or numbers
All options are made as short as possible and there is no intention to 
try to make the option name
understandable, but in the documentation they should all be explained. A 
given designation for an
option should mean the same in all commands if possible.
e.g. /m   is always allow missing values

set:
a general specification for running the programme or setting formats. 
The first word of a "set"
tells what this specifices. General which cannot obey this rule should 
be avoided.
e.g. set display databrowser = on/off   (will always show data browser 
in the background).
.................................................................................................................
All showing of labels, values etc are defined by these "set"  commands:
 set var label = on/off  ( show or hide variable label)
 set var name = on/off    (removes name of var if the same as first word 
in variable label)
 set var value = on/off  (controls whether actual numerical value are 
shown)
 set var label = on/off (controls showing of "comment legal/value labels")

e.g. set var value = off: show only value labels in tables (but show 
values if no value labels are defined)
.................................................................................................................
Creation and display of statistical tests, percentages etc are moved to 
each command:
e.g.
means age sex          (only descriptive summaries are shown)
means age sex    /t    (will display t test if one group and f test 
+Bartletts test with more groups)
tables agegrp sex     (only showns counts)
tab agegrp sex     /chi   (shows chi square values)
tab agegrp sex   /r  /c    (would add row and column percentages)
tab agegrp sex /t     (would add total percentages with no decimal points)
.................................................................................................................
Format of estimates and confidence intervals and table column headers 
are defined as "set":
e.g.
TABLE PERCENT HEADER ROW    %
TABLE PERCENT FORMAT ROW    P1{}
.................................................................................................................
Format of Confidence Intervals are

TABLE CI HEADER    (95% CI)
TABLE CI FORMAT    C2-()
.................................................................................................................

Regards

Jens Lauritsen
EpiData Association


More information about the EpiData-list mailing list